Automated cell counting for Trypan blue-stained cell cultures using machine learning.

Area of Science:

• Biotechnology
• Machine Learning Applications
• Cell Biology

Background:

• Accurate cell counting is essential for cell culture maintenance, viability assessment, and proliferation rate determination.
• Manual cell counting is time-consuming, especially for parallel cultures, and automated counters are expensive.
• There is a need for efficient, accurate, and cost-effective cell counting methods.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for accurate and rapid cell counting.
• To provide a cost-effective alternative to existing automated cell counting solutions.
• To assess the model's performance on various cell lines and its suitability for high-throughput analysis.

Main Methods:

• A YOLOv4-based machine learning model was trained and validated using images of Trypan blue-stained insect cell lines (Trichoplusia ni and Spodoptera frugiperda).
• The model's accuracy was evaluated using F1 scores for both live and dead cells.
• The model was further tested on unstained human embryonic kidney (HEK) cells to assess its versatility.

Main Results:

• The ML model achieved high accuracy, with F1 scores of 0.97 for alive and 0.96 for dead cells.
• The model demonstrated versatility, achieving an F1 score of 0.96 on untrained human embryonic kidney cells.
• The implementation features a user-friendly interface and batch processing capabilities, suitable for high-throughput experiments.

Conclusions:

• The developed ML model provides a fast, accurate, and bias-free method for cell counting.
• This approach offers a significant improvement over manual counting and a cost-effective alternative to expensive automated counters.
• The model's versatility makes it applicable to diverse cell lines and research contexts, including experimental design evaluations.